Core: Introduce CompositeMetricsReporter

[nastra]

This introduces a CompositeMetricsReporter that allows registering multiple (simple) MetricsReporter instances that all get notified whenever a new MetricsReport is available.

This also switches from a collection of reporters to the CompositeMetricsReporter in TableScanContext to fix an issue where duplicate reporters were not handled (see #7337 (comment) for details).

[nastra]

@rdblue as we discussed I've made this an inner class. However, we could consider moving the class out of CatalogUtil and replace https://github.com/apache/iceberg/blob/master/core/src/main/java/org/apache/iceberg/TableScanContext.java#L359-L366 with a CompositeMetricsReporter.

For cases where we'd like to use multiple reporters during scans

TableScan tableScan = table
        .newScan()
        .metricsReporter(...)
        .metricsReporter(...)
        .metricsReporter(...);

we would internally instantiate a CompositeMetricsReporter in TableScanContext#reportWith(..) and then have TableScanContext#metricsReporter() return MetricsReporter rather than Collection<MetricsReporter>.

[nastra]

In fact I just realized that the current implementation of TableScanContext#reportWith() that uses Collection<MetricsReporter> actually can contain the same reporter multiple times (which I've seen in some tests where the same thing is reported twice to the log file)

[rdblue]

Why do we need to support multiple reporters? I don't think we need to change the behavior.

[rdblue]

I think this should be RuntimeException since report doesn't throw a checked exception.

[nastra]

the reason I'm catching an Exception here is that I think we generally don't want to interfere with actual scans or commits when anything on the reporting side fails. Generally speaking, every reporter should make sure that it doesn't interefere with the scan/commit path when reporting and something fails but I'm not sure we can enforce that.
I'm a bit worried that if we only catch RuntimeException and some reporter misbehaves in an unexpected way, we'd have to go back and fix either the misbehaving reporter or this catch clause.
thoughts?

[rdblue]

Looks good. Just one minor thing to fix with the sets.